Embed Size (px)
Citation preview
DI
SC
US
SI
ON
P
AP
ER
S
ER
IE
S
Forschungsinstitut zur Zukunft der ArbeitInstitute for the Study of Labor
The Academic Impact of Natural Disasters:Evidence from L’Aquila Earthquake
IZA DP No. 8867
February 2015
Giorgio Di Pietro
The Academic Impact of Natural Disasters:
Evidence from L’Aquila Earthquake
Giorgio Di Pietro Westminster Business School
and IZA
Discussion Paper No. 8867 February 2015
IZA
P.O. Box 7240 53072 Bonn
Germany
Phone: +49-228-3894-0 Fax: +49-228-3894-180
E-mail: [email protected]
Any opinions expressed here are those of the author(s) and not those of IZA. Research published in this series may include views on policy, but the institute itself takes no institutional policy positions. The IZA research network is committed to the IZA Guiding Principles of Research Integrity. The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center and a place of communication between science, politics and business. IZA is an independent nonprofit organization supported by Deutsche Post Foundation. The center is associated with the University of Bonn and offers a stimulating research environment through its international network, workshops and conferences, data service, project support, research visits and doctoral program. IZA engages in (i) original and internationally competitive research in all fields of labor economics, (ii) development of policy concepts, and (iii) dissemination of research results and concepts to the interested public. IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper should account for its provisional character. A revised version may be available directly from the author.
IZA Discussion Paper No. 8867 February 2015
ABSTRACT
The Academic Impact of Natural Disasters: Evidence from L’Aquila Earthquake
This paper examines the effect of the L’Aquila earthquake on the academic performance of the students of the local university. Following this traumatic event, not only are students likely to have developed an acute stress disorder, but they have also experienced a significant disruption in their learning environment because of the closure of several university buildings and the relocation to temporary premises. We compare changes in educational outcomes among students of the University of L’Aquila before and after the earthquake with changes in educational outcomes during the same time period among students enrolled at other Central Italian universities. The empirical results suggest that while this natural disaster has reduced students’ probability of graduating on time by 6.6 percentage points, it has had no statistically significant effect on university drop-out in the very short-term. Additionally, the on-time graduation result masks differential gender effects. JEL Classification: Q54, I23 Keywords: L’Aquila earthquake, academic performance, difference-in-differences Corresponding author: Giorgio Di Pietro University of Westminster Westminster Business School Department of Economics and Quantitative Methods 35 Marylebone Road NW1 5LS London United Kingdom E-mail: [email protected]
3
1. Introduction
During the night of 6th April 2009 the city of L’Aquila was struck by a violent earthquake
that killed 309 individuals and injured about 1,600 people. Thousands of houses and
buildings were seriously damaged or destroyed. About 90 percent of the residents of L'Aquila
were displaced from their homes following this natural disaster. The mediaeval centre of
L’Aquila, which hosts the local University, was especially hit hard. This event was one of the
most expensive natural disasters in Italy’s history with an estimated total economic impact of
about 540 million euro (Commissariato delegato per la ricostruzione in Abruzzo, 2011).
Many accounts of the event and its aftermath also received worldwide media coverage thanks
to the fact that the 35th G8 international summit was held in L’Aquila in July 2009.
This paper examines the impact that the earthquake had on the academic performance of the
students of University of L’Aquila. Such an event had significant disruptions to their daily
life. Housing was an immediate concern. A large number of them had to relocate to cities and
villages miles away from L’Aquila. Many students moved to Avezzano, which is situated
about 31 miles from L’Aquila. Additionally, about 70 percent of the infrastructure of the
University of L’Aquila (including university buildings, libraries and student canteens) was
closed following the earthquake (Maggiolo, 2010). Though several temporary locations were
found to enable students to continue their academic studies, the quality of educational
programs inevitably suffered as a result of the earthquake. There was also a transportation
problem as students had to travel to the new university locations. Despite efforts to return to
the pre-earthquake situation, the pattern of disruption at the university continued for several
years after the natural disaster. For instance, students of the Faculty of Engineering were able
to return to their original university campus only in October 2013.
Perhaps more importantly, many students of the University of L’Aquila are likely to have
experienced psychological problems in the aftermath of this natural catastrophe. There are
4
many papers indicating the presence of post-traumatic stress disorder (PTSD) in survivors of
natural disasters (see, for instance, Rowe and Liddle, 2008; Miller, 2005)1. PTSD symptoms
include poor concentration, depression, anxiety and insomnia. In addition to PTSD, exposure
to natural disasters may increase the risk of developing health issues such as cardiorespiratory
and musculoskeletal problems (see, for instance, Keskinen-Rosenqvist et al., 2011). Studies
on the effects of the L’Aquila earthquake are consistent with the above findings. For instance,
a recent paper (Di Castelnuovo et al., 2013) compares the health status of residents of
L’Aquila with the health status of people living in another region of Central Italy (i.e. Molise)
in the 6 months following the earthquake. It is found that the proportion of people suffering
from metabolic syndrome2 (MS) is significantly higher among the former relative to the
latter. Another study (Tempesta et al., 2013) looks at differences in sleep quality among
L’Aquila citizens before and after the earthquake. The results indicate that exposure to the
natural disaster resulted in a significant deterioration of sleep quality and increased incidence
of disruptive nocturnal behaviours (DNB). Furthermore, it is possible that the earthquake had
a profound emotional impact especially on the University of L’Aquila’s student community.
Ten of the victims of the earthquake were students who, during the night of the earthquake,
were at a dormitory located in the historical centre of L’Aquila.
While conventional wisdom suggests that a disruption in the learning environment caused by
natural disasters negatively affects students' academic achievement, several papers (see, for
instance, Smilde-van den Doel, et al., 2006) indicate that the direction and the magnitude of
this effect are complex to evaluate as they depend on various factors. In particular,
interventions in favour of the affected students in the aftermath of natural catastrophes can
play an important role in this context. Students may not experience a performance decline or 1 Some papers (e.g. Carter et al., 2014) find evidence of these problems among university students. 2 Metabolic syndrome is a cluster of conditions — increased blood pressure, high blood sugar level, excess body fat around the waist and abnormal cholesterol levels — that occur together, increasing a person’s risk of heart disease, stroke and diabetes.
5
may experience a lesser decline if measures are taken to help them deal with post-natural
disaster issues.
Several such measures were taken shortly after the L’Aquila earthquake in an attempt to
mitigate some of the negative effects that this event had on students. Students were given a
fee exemption for three academic years3. They were also entitled to receive discounts on
study-related equipment like textbooks and computers. Additionally, to help students address
transportation issues caused by relocation, they benefited from free public transport. Finally,
more scholarships were awarded to students thanks to the donations of several national and
international institutions.
We focus our attention on the impact of the L’Aquila earthquake on university drop-out and
on-time graduation (i.e. within 3 years after initial enrolment). These are very important
educational attainment measures in Italy where a large proportion of university students
abandon their studies before graduation (Schnepf, 2014) and many of those who manage to
complete a degree do not do this within the prescribed time (Aina et al., 2011).
Our empirical strategy is to examine the differences in drop-out and on-time graduation
between students enrolled at the University of L’Aquila (the treated group) and those enrolled
at other universities located in central Italy (the control group)4, and then look at how these
differences changed in the post-earthquake period relative to the pre-earthquake period. This
comparison, which generates a difference-in-differences (DiD), allows us to control for time-
varying unobservables affecting drop-out and on-time graduation across students at both the
University of L’Aquila and other central Italian universities. This strategy hinges on the
assumption that in the absence of the earthquake the difference in our educational outcomes 3 The fee exemption was successively extended for three more academic years. 4 Students enrolled at universities located in the regions of Molise, Marche and Abruzzi (apart from L’Aquila) comprise the control group. The dataset used does not include the specific name of the university, but does provide information about the province where the university is located.
6
between students of the University of L’Aquila and those enrolled at other central Italian
universities would have been the same as the one in the pre-earthquake period.
Students of other universities located in central Italy constitute a credible control group for
three reasons. First, as reported in Section 4, these students are similar to those of the
University of L’Aquila in most observable characteristics. Second, as shown in Section 5, in
the pre-earthquake period, conditional on controls, the trend in on-time graduation across
students at the University of L’Aquila is similar to that across students at other central Italian
universities. Third, geographical proximity implies that treated and control students are
embedded in the same socio-cultural-economic environment. For example, the cost of living
of these students is similar and they also tend to share similar cultural traits.
The remainder of the paper is as follows. Section 2 reviews the findings of those studies
investigating the effects of natural disasters on the academic performance of university
students. Section 3 outlines the empirical strategy, while Section 4 describes the data. Section
5 presents and discusses the results of the empirical analysis. Section 6 concludes.
2. Literature review
A relatively large number of studies investigate the impact of natural disasters on students’
progression and attainment. However, only a few of them examine the impact that these
events had on the academic performance of university students. Two papers study how
earthquakes have affected the achievement of students at Canterbury University, New
Zealand. The first one is by Helton et al. (2011) and looks at the effect of an earthquake that
occurred in 2010. It finds that this natural disaster had a negative effect on cognition.
However, one needs to be cautious about the conclusion of this study as it is based on a small
sample (i.e. 18 students), participants are self-selected and no control group was used. Some
of these problems are addressed in the second paper (Wilkinson et al., 2013) that compares
7
the disruptions caused by the 2010 earthquake with those related to an earthquake that took
place in 2011. Focusing on medical students, the authors find that the former had a greater
impact on assessment performance relative to the latter. An explanation for this result is that,
while the 2011 earthquake happened at the beginning of the academic year therefore giving
students and the institution time to develop adaptive or coping strategies before exams, this
did not occur with the 2010 earthquake, which took place at the end of the academic year.
Krane et al. (2007) analyse the disruptions caused by Hurricane Katrina to Tulane University
School of Medicine and Louisiana State University School of Medicine at New Orleans.
Following this event, these Schools were relocated to temporary locations and training for
students was re-established one month later. Hurricane Katrina seems to have had no effect
on student performance. Mean final course average, standardized National Board of Medical
Examiners course examination scores and scores for US Medical Licensing Examination
Steps 1 and 2 were all unchanged. This result, however, contrasts with that of Watson et al.
(2011,) who investigate the effect of Hurricane Ilke on students of the University of Texas
Medical Branch (UTMB). Using a representative sample of the UTMB student population,
the authors find that about half of the respondents reported that this natural disaster had
negatively affected their academic performance. A significant proportion (24%) of the
students interviewed considered that the size of the negative effect was ‘substantial’.
In sum, the findings of these studies indicate that, though natural disasters have the potential
to exert a considerable negative effect on student academic achievement, this does not always
happen. Some institutions are able to handle the disruptions of the disaster relatively well,
thereby mitigating or even completely offsetting the negative consequences on students'
performance.
8
3. Empirical strategy
We pool data from 3 cohorts of students who enrolled at universities located in central Italy
in 2001, 2004 and 2007. Data allow us to follow these students during their first 3/4 years of
study. In order to evaluate the effect of the L’Aquila earthquake on drop-out and on-time
graduation, we use a DiD strategy. The first difference contrasts students of the University of
L’Aquila and students enrolled at other universities of central Italy, assuming therefore that
the latter were unaffected by the earthquake. The second difference concerns the timing of
the earthquake. In our data, while students of the University of L’Aquila who started their
studies in 2007 were hit by this natural disaster, those who enrolled at the same institution in
2001 and 2004 were not exposed to the earthquake. The difference in these differences can be
interpreted as the causal effect of the earthquake, under the assumption that in the absence of
the earthquake, the change in our outcome measures would have not been systematically
different across students at the University of L’Aquila and those at other universities in
central Italy.
The following model is estimated:
ijttjtjijtijt EarthquakeAquilaEarthquakeAquilaXY εβββββ +++++= *432'
10 (1)
where ijtY is a dichotomous variable that is equal to 1 if student i enrolled at university j in
year t drops out /graduates on time, and 0 otherwise. X is a vector of individual and family
characteristics that are thought to influence these outcomes. Aquila is a dummy variable that
takes the value 1 if the student was enrolled at the University of L’Aquila, and 0 otherwise.
Earthquake is a dummy taking the value 1 if the student started university in 2007, and 0
otherwise; ε is the error term.
9
The coefficients of this DiD framework have the following interpretation. 2β captures the
average permanent differences in the outcomes between students enrolled at the University
of L’Aquila and those enrolled at other central Italian universities . 3β accounts for pre and
post-earthquake differences that are common to both the treated and control groups. Our
parameter of interest is 4β . It tests whether the difference in average outcomes between
students enrolled at the University of L’Aquila and those enrolled at other central Italian
universities in the post-earthquake period is different from the same difference in the pre-
earthquake period. The key identifying assumption is that 4β would be zero in the absence of
the L’Aquila earthquake. 4β identifies the true impact of the earthquake assuming that,
conditional on X , trends student in drop-out and on-time graduation between the University
of L’Aquila and other central Italian universities would be identical if the earthquake had no
occurred.
Although the dependent variable of Equation (1) is dichotomous, this equation can be
estimated by ordinary least squares (OLS) as this allows us to estimate the parameter of
interest directly5. As pointed out by Angrist (2001), the problem of causal inference does not
significantly differ between limited dependent variables and continuous outcomes. This
means that if there are no covariates or the covariates are sparse and discrete, then linear
models can be used to estimate models with limited dependent variables as well as models
with other types of dependent variables.
4. Data
Data come from three waves (i.e. 2004, 2007 and 2011) of a national cross-sectional survey
(Percorsi di studio e di lavoro dei diplomati) conducted by the Italian National Statistical
5 As shown by Ai and Norton (2003), the interpretation of the marginal effect of an interaction term in nonlinear models can be problematic and in some cases it can lead to a misleading conclusion.
10
Institute (ISTAT). Each wave consists of a nationally representative sample of high school
leavers who were surveyed 3/4 years after successfully completing their studies. As one of
the possible destinations of recent high school leavers is university enrolment6, these data
enable us to identify three cohorts of students who began university in 2001 (from the 2004
wave), 2004 (from the 2007 wave) and 2007 (from the 2011 wave). As the earthquake
occurred in April 2009, while the 2001 and 2004 cohorts were not exposed to this natural
disaster, those students of the University of L’Aquila who began their studies in 2007 were
affected by it. More precisely, they were hit by the earthquake while they were in their
second year of study7.
We employ two measures of student outcome in this study. The first is university drop-out.
Unfortunately, for this analysis we can only use data from the 2007 and 2011 waves. While
the 2004 wave does not provide information on the university attended for those students who
have dropped out, such information is reported in later waves. Our second measure is whether
or not the student has successfully completed a First-level degree. Following the European
framework of the Bologna Process, in 2001 the Italian higher education system embarked on
a process of reform that led to the introduction of the so-called ‘3+2’ model. This consists of
a First-level degree (Laurea di primo livello) that lasts three years, followed by a Second-
level degree (Laurea specialistica) of two years length. For this outcome, data from all three
waves can be employed as the 2004 wave does give information on the university attended
for those students who completed a First-level degree or were still enrolled at the time of the
survey.
6 In Italy, all individuals who have successfully completed five years at any type of upper secondary school gain the automatic right to enrol at university. 7 In Italy, the academic year runs between the beginning of October and the end of July the following year.
11
One issue with merging data from the three waves of this survey is that, while in the 2004
and 2007 waves high school leavers were surveyed three years after successfully completing
their studies, in the 2011 wave students were contacted four years following the end of high
school. This poses a problem in creating harmonised outcome measures as university students
need to be observed for the same amount of time across cohorts. Such a problem can be
easily solved with respect to our second measure of student outcome as the 2011 wave
reports information on the year students successfully completed the First-level degree.
Therefore, in all three cohorts we can observe whether the student earned a First-level degree
within three years following his/her initial enrolment. However, what of the drop-out
indicator? Given the purpose of this study, our only viable option is to see whether the
student has withdrawn from university at the end of his/her second year of study. While
information on third-year drop-out is available in the 2004 cohort, in the 2007 cohort data do
not allow us to separate third-year drop-out from fourth-year drop-out. Second-year drop-out
captures the immediate effect of the earthquake on drop-out since, as mentioned above,
students of the University of L’Aquila were in their second year of study when the natural
disaster occurred.
The survey contains information on many individual characteristics that are likely to
influence the probability of dropping out of university as well the probability of earning a
First-level degree within three years after initial enrolment. We are therefore able to control
for gender, age, region of residence8, type of upper secondary school (academic versus
technical and private versus public), academic ability (proxied by lower and upper secondary
school final marks) and parental education. The unemployment rate (among people aged
between 20 and 24 and broken down by province) is also included in our specification in an
8 Though we focus our attention only on students enrolled at universities located in three regions, several of these students indicated their residence in many other different Italian regions. Some of the students may have been commuting. Additionally, university attendance is generally not compulsory in Italy.
12
attempt to study the impact of local labour market conditions on our student outcome
measures.
To construct the final samples, we eliminated those high school leavers who did not enrol at
the university within 3/4 years following successful completion of their studies, those high
school leavers who enrolled at university but did not do so straight after the conclusion of
their studies9, those high school leavers who were living abroad at the time of the survey and
those high school leavers with missing information of interest. Table 1 reports the summary
statistics for our final samples. The sample used to investigate the effect of the L’Aquila
earthquake on on-time graduation comprises 2,570 students. Among those students who were
still enrolled at university after 3 years, 8.9 percent obtained a First-level degree. On the other
hand, our drop-out sample, which only includes students from the 2004 and 2007 cohorts,
consists of 2,041 individuals. Conditional on not having dropped out previously, the
proportion of university students who abandoned their studies in the second year is 2.4
percent.
Insert Table 1 about here
Table 2 presents summary statistics for students of the University of L’Aquila and those of
other Central Italian universities in the pre-earthquake period. While for the drop-out analysis
this period includes only the 2007 cohort, it comprises both the 2004 and 2007 cohorts for the
on-time graduation analysis. In the latter, the only relevant statistically significant differences
(at 1% and 5%) between the treated and control groups are some dummy variables related to
lower and upper secondary school final marks. Looking at the drop-out sample, in the pre-
disaster period the proportion of students from academic upper secondary school (liceo) is
9 Students who went to university one or more years following high school completion are observed for a shorter period of time relative to those who began university immediately after the end of high school. In each wave the large majority of high school leavers (i.e. more than 85%) started university straight after successfully completing their school studies.
13
found to be larger at the University of L’Aquila relative to other Central Italian universities.
The only other relevant statistically significant difference is the proportion of students who
achieved “good” as a lower secondary school final grade.
Insert Table 2 about here
5. Results
Table 3 presents the DiD estimates from Equation (1), which examines the effect that the
L’Aquila earthquake had on our two university performance measures. While the first half of
Table 3 reports the on-time graduation results, the second half of Table 3 presents the drop-
out findings. In both regressions we include all our control variables.
Insert Table 3 about here
Academic ability is, as expected, an important predictor of the probability of obtaining a
First-level degree within 3 years after initial enrolment. Performing very well both at upper
and lower secondary school significantly increases the probability of graduating on time.
There is also a gender effect, as women are found to be 2.9 percentage points more likely to
graduate on time relative to men.
Turning to the primary purpose of this paper, it should be noted that the DiD coefficient has a
negative sign and is statistically significant at the 5% level. Our result indicates that the
change in the probability of on-time graduation between L’Aquila students and those enrolled
at other central Italy universities is about 6.6 percentage points lower in the post-earthquake
period relative to the pre-earthquake period.
Moving to the second half of Table 3, our estimates are seen to be broadly in line with those
obtained by previous studies (see, for instance, Di Pietro, 2006). Having studied at an
14
academic track upper secondary school (liceo) decreases the likelihood of dropping out of
university. The pattern of coefficients on age groups is monotonic, so that drop-out is highest,
other things being equal, for students in the highest age category. Unsurprisingly, students
with a higher final grade at upper secondary school are found to have a lower propensity to
abandon university, ceteris paribus.
In the second half of Table 3, the DiD coefficient has a positive sign as expected but is
statistically indistinguishable from zero (p=0.125). A potential explanation why the
earthquake did not have any immediate effect on university drop-out is that, following the
natural disaster, all students of the University of L’Aquila received a fee exemption for the
next 3 academic years. In light of this, students may have had no interest in formally
abandoning their studies in the academic year 2009/2010 even if, after the earthquake, they
were reluctant to continue studying at the university. As they did not have to pay any fee,
these students may have taken some time before taking the decision to drop out of university.
Additionally, even though we include local unemployment rate among the explanatory
variables of our model, perceived future unfavourable labour market conditions could have
discouraged students of the University of L’Aquila from abandoning their studies.
Worsening labour market prospects reduces the opportunity cost of studying thereby
increasing the relative attractiveness of university education.
Two different tests are carried out to check the validity of our on-time graduation DiD results.
First, we look at the identifying assumption of parallel trend in this outcome in the absence of
the earthquake10. A similar trend in student on-time graduation across the University of
L’Aquila and other universities located in central Italy should be observed prior to the
earthquake. If this assumption is violated, then our estimated effect of the earthquake might 10 Unfortunately, the parallel trend assumption cannot be checked for our drop-out analysis as we only have data on one student cohort before the earthquake. For the same reason no placebo test can be performed for the drop-out analysis
15
be spurious as it could be the result of a diverging trend in on-time graduation that existed
even before the occurrence of the natural disaster. To test the parallel trend assumption,
Equation (1) is re-estimated using data only for the 2001 and 2004 student cohorts. The
results, which are shown in the first half of Table 4, indicate that in the pre-earthquake period
the student on-time graduation trend was similar across the treated and untreated universities
as the coefficient on the relevant interaction term is statistically insignificant.
Insert Table 4 about here
Second, we perform a placebo test assuming that the earthquake affected students of the
University of L’Aquila who began their studies in 2004 instead of those who enrolled at this
institution in 2007. For this falsification exercise, we chose our middle cohort because it
corresponds to the mid-point of the pre-earthquake period. Results are quite similar if
students of the University of L’Aquila in the 2001 cohort are the ones artificially assumed to
have been exposed to the earthquake. The second half of Table 4 reports results from the
corresponding DiD model that is based on this mis-coded earthquake dummy. The DiD
coefficient is found not to be statistically significant at conventional levels. Such a finding
indicates that the estimated effect found in the first half of Table 3 is a genuine causal effect
of the earthquake and not a an artefact of the statistical approach.
Next, we examine whether the effect of the L’Aquila earthquake on academic performance
varies by gender. Given that several studies (see, for instance, Cairo et al., 2010; Hashmi, et
al., 2011; Zhang, et al., 2011) conclude that women are more likely to suffer from PTSD after
an earthquake than men, it is possible that the L’Aquila earthquake had a detrimental effect
especially on the academic performance of women. To test this hypothesis, we estimate our
DiD model separately for men and women. While the first half of Table 5 shows the on-time
graduation findings, the second half of Table 5 reports the drop-out estimates. Our estimates
16
show that the effect triggered by the earthquake on the probability of on-time graduation
differs across gender. Although, as expected, the estimated DiD coefficient has a negative
sign for both sexes, not only is its size larger for women but is also statistically significant
only for women. Specifically, our result implies an 11.9 percentage point decline in the
probability of women earning a First-level degree in three years. This finding suggests that
the overall effect found in the first half of Table 3 is basically driven by women.
Insert Table 5 about here
Estimates reported in the second half of Table 5 indicate that the L’Aquila earthquake had no
immediate effect on university drop-out neither for men nor for women. Though the DiD
coefficients have the expected positive sign, none of them is found to be statistically
significant at conventional levels.
6. Conclusions
On 6th April 2009 the city of L’Aquila was hit by a violent earthquake that caused hundreds
of victims and considerable damage to buildings, road and other infrastructures. In this paper,
we investigate the impact that this natural disaster had on the academic performance of the
students of the local university. Not only has the earthquake led to significant disruptions in
the learning environment (e.g. closure of university buildings), but students of the University
of L’Aquila are likely to have suffered from physiological and psychological stress following
this event.
In order to evaluate the effect of the earthquake on student academic performance, we use a
difference-in-differences methodology. Changes in outcomes for students at the University of
L’Aquila before and after the earthquake are compared to changes in outcomes for students at
other central Italian universities over the same period of time. The key identifying assumption
17
of this approach is that trends in student outcomes across these two groups of universities
would have been identical in the absence of the earthquake.
The empirical results suggest that the L’Aquila earthquake has significantly reduced the
probability that a student will graduate on time. Specifically, our estimates indicate that this
natural disaster led to a decline of about 6.6 percentage points in the likelihood of earning a
First-level degree within three years of initial enrolment. On the other hand, we find no
evidence that the L’Aquila earthquake had any statistically significant effect on university
drop-out during the academic year following this natural disaster. This finding could be
explained by perceived unfavourable future labour market conditions as well as by the fee
exemption policy introduced by the University of L’Aquila following the earthquake.
Our results also show that the overall negative effect exerted by the L’Aquila earthquake on
the probability of graduating on time masks significant differences across gender. While this
natural disaster has considerably decreased the likelihood that female students of the
University of L’Aquila will graduate in three years, no corresponding statistically significant
effect is found for male students. This finding is consistent with the conclusions of several
papers according to which, following an earthquake, women, relative to men, are more likely
to experience mental problems such as depression, anxiety and insomnia.
18
References
Ai, C. and Norton, E. (2003) “Interaction terms in logit and probit models”, Economics
Letters 80(1), pp. 123-129.
Aina, C., Baici, E. and Casalone, G. (2011) "Time to degree: students' abilities, university
characteristics or something else? Evidence from Italy", Education Economics 19(3), pp.
311-325.
Angrist, J.D. (2001) “Estimation of a limited dependent variable models with dummy
endogenous regressors: Simple strategies for empirical practice”, Journal of Business and
Economic Statistics 19(1), pp. 2-16.
Cairo, J.B., Dutta, S., Nawaz, H., Hashmi, S., Kasl, S. and Bellido, E. (2010) “The
prevalence of post-traumatic stress disorder among adult earthquake survivors in Peru”,
Disaster Med Public Health Preparedness 4(1), pp. 39-46.
Carter, F.A., Bell, C.J., Ali, A.N., McKenzie, J. and Wilkinson, T.J. (2014) “The impact of
major earthquakes on the psychological functioning of medical students: a Christchurch, New
Zealand study”, The New Zealand Medical Journal 127(1398), pp. 54-66.
Commissariato delegato per la ricostruzione in Abruzzo (2011) Prime considerazioni sugli
scenari di sviluppo e le strategie di intervento da mettere in campo per il territorio
dell’Aquila, available at: http://www.commissarioperlaricostruzione.it/Informare/Normative-
e-Documenti/Atti-e-documenti-della-Struttura-Tecnica-di-Missione-STM/Idee-e-strumenti-
per-la-ricostruzione-pesante/Prime-considerazioni-sulle-strategie-di-sviluppo-e-gli-scenari-
di-intervento-da-mettere-in-campo-per-il-territorio-dell-Aquila, accessed 2/2/2015.
19
Di Castelnuovo, A., Di Pietro, N., Di Tomo, P., Di Silvestre, S., Pipino, C., Nenna, G.,
Bonomini, M., Iacoviello, L. and Pandolfi ,A. (2013) “Metabolic syndrome in survivors from
the 2009 earthquake in Italy”, Nutrition Metabolism and Cardiovascular Diseases 23(1),
pp.5-8.
Di Pietro, G. (2006) “Regional labour market conditions and university dropout rates:
evidence from Italy”, Regional Studies 40(6), pp. 617-630.
Hashmi, S., Petraro, P., Rizzo, T., Nawas, H., Choudlhary, R., Tessier-Sherman, B., Kasl and
Nawaz, H. (2011) “Symptoms of anxiety, depression and post-traumatic stress among
survivors of the 2005 Pakistani Earthquake”, Disaster Med Public Health Preparedness 5(4),
pp. 293-299.
Helton, W. S., Head, J. and Kemp, S. (2011) “Natural disaster induced cognitive disruption:
Impacts on action slips”, Consciousness and Cognition 20(4), pp. 1732-1737.
Keskinen-Rosenqvist, R., Michélsen, H., Schulman, A. and Wahlström, L. (2011) “Physical
symptoms 14 months after a natural disaster in individuals with and without injury are
associated with different type of exposure”, Journal of psychosomatic research 71(3), pp.
180-187.
Krane, N.K., DiCarlo, R.P. and Kahn, M.J. (2007) “Medical Education in Post-Katrina New
Orleans. A story of survival and renewal”, JAMA The Journal of the American Medical
Association 298(9), pp.1052–1055.
Maggiolo, A. (2010) Aprile 2009 - Aprile 2010: l'Università dell'Aquila un anno dopo il
terremoto, Studentiit, 18 March, Available at:
20
http://www.studenti.it/universita/inchieste/universita-aquila-un-anno-dopo-il-terremoto.php,
accessed 2/2/2015.
Miller, G. (2005)” The Tsunami’s psychological aftermath”, Science 425, 1030-1033.
Rowe, C.L. and Liddle, H.A. (2008) “When the levee breaks: treating adolescents and
families in the aftermath of hurricane Katrina”, Journal of Marital and Family Therapy 34,
pp. 132-148.
Schnepf, S. V. (2014) Do tertiary dropout students really not succeed in European labour
markets?, IZA Discussion Papers 8015, IZA, Bonn.
Smilde-van den Doel, D.A., Smit, C. and Wolleswinkel- van den Bosh, J.H. (2006) “Before
and after a disaster school performance and social-emotional behaviour of primary school
children”, Paediatrics, 118(5), pp. 1311-1320.
Tempesta, D., Curcio, G., De Gennaro, L. and Ferrara, M. (2013) “Long-Term Impact of
Earthquakes on Sleep Quality”, PLoS ONE 8(2), e55936.
Watson, P.G., Loffredo, V.J. and Mckee J.C. (2011) “When a natural disaster occurs: lessons
learned in meeting students’ need”, Journal of Professional Nursing 6, pp 362-369.
Wilkinson, T. J., Ali, A.N. , Bell, C.J., Carter, F.A., Frampton, C.M. and McKenzie, J.M.
(2013) “The impact of learning environment disruption on medical student performance”,
Medical education 47(2), pp. 210-213.
Zhang, Z., Shi, Z., Wang L. and Liu, M. (2011) “One year later: mental health problems
among survivors in hard hit areas of the Wenchuan earthquake”, Public Health 125(5), pp.
293-300.
21
Table 1: Summary statistics
On-time graduation sample
Drop-out sample
Mean Std. Dev. Mean Std. Dev. Graduated on-time (conditional on still being enrolled at university after 3 years)
0.089 0.285
Dropped-out in the second year (conditional on not having dropped out in the first year)
0.024 0.154
Aquila (Enrolled at the University of L’Aquila)
0.145 0.352 0.166 0.372
male 0.371 0.483 0.385 0.487 Age -21 years or less 0.046 0.209 0.039 0.193 -22 years 0.847 0.360 0.866 0.341 -23 years or more 0.107 0.309 0.096 0.294 Upper secondary school final mark
- between 60 and 69 0.191 0.393 0.197 0.398 -between 70 and 79 0.267 0.443 0.254 0.435 -between 80 and 89 0.205 0.404 0.205 0.404 -between 90 and 100 0.337 0.473 0.345 0.475 Lower secondary school final mark
-pass 0.114 0.318 0.104 0.305 -good 0.271 0.445 0.290 0.454 -very good 0.292 0.455 0.302 0.459 -excellent 0.323 0.468 0.304 0.460 Academic upper secondary school (liceo)
0.479 0.500 0.450 0.498
Private upper secondary school
0.026 0.160 0.032 0.176
Father’s highest education is university degree
0.148 0.355 0.145 0.352
Mother’s highest education is university degree
0.141 0.348 0.130 0.336
Unemployment rate 17.066 7.021 15.583 6.038 Number of observations 2,570 2,041 Sampling weights used
22
Table 2: Pre-earthquake summary statistics for the treated and untreated On-time graduation sample Drop-out sample University
of L’Aquila
Other central Italian
universities
Diff University of
L’Aquila
Other central Italian
universities
Diff
Graduated on-time (conditional on still being enrolled at university after 3 years)
0.053 (0.225)
0.070 (0.255)
-0.017 (0.18)
Dropped-out in the second year (conditional on not having dropped out in the first year)
0.025 (0.156)
0.024 (0.154)
-0.001 (0.013)
male 0.378 (0.486)
0.359 (0.480)
0.019 (0.034)
0.350 (0.478)
0.375 (0.484)
-0.025 (0.040)
Age -21 years or less 0.045
(0.209) 0.051
(0.221) -0.006 (0.016)
0.041 (0.199)
0.046 (0.210)
-0.005 (0.017)
-22 years 0.822 (0.384)
0.832 (0.374)
-0.010 (0.027)
0.835 (0.372)
0.848 (0.359)
-0.013 (0.030)
-23 years or more 0.133 (0.340)
0.117 (0.322)
0.016 (0.023)
0.124 (0.330)
0.106 (0.308)
0.018 (0.026)
Upper secondary school final mark
- between 60 and 69) 0.256 (0.438)
0.177 (0.381)
0.079*** (0.028)
0.137 (0.345)
0.198 (0.399)
-0.061* (0.032)
-between 70 and 79 0.235 (0.425)
0.285 (0.452)
-0.050 (0.032)
0.245 (0.431)
0.270 (0.444)
-0.025 (0.037)
-between 80 and 89 0.222 (0.417)
0.193 (0.395)
0.029 (0.029)
0.201 (0.402)
0.184 (0.387)
0.017 (0.032)
-between 90 and 100 0.286 (0.453)
0.345 (0.476)
-0.059* (0.034)
0.418 (0.495)
0.348 (0.477)
0.070* (0.040)
Lower secondary school final mark
-pass 0.123 (0.330)
0.131 (0.337)
-0.008 (0.024)
0.139 (0.347)
0.123 (0.329)
0.016 (0.027)
-good 0.354 (0.479)
0.263 (0.440)
0.091*** (0.032)
0.248 (0.433)
0.325 (0.469)
-0.077** (0.038)
-very good 0.292 (0.456)
0.283 (0.450)
0.009 (0.032)
0.293 (0.457)
0.295 (0.456)
-0.002 (0.038)
-excellent 0.230 (0.422)
0.324 (0.468)
-0.094*** (0.033)
0.321 (0.468)
0.256 (0.437)
0.065* (0.037)
Academic upper secondary school (liceo)
0.475 (0.501)
0.477 (0.500)
-0.002 (0.036)
0.526 (0.501)
0.410 (0.492)
0.116*** (0.041)
Private upper secondary school 0.034 (0.181)
0.025 (0.158)
0.009 (0.012)
0.023 (0.149)
0.039 (0.194)
-0.016 (0.016)
Father’s highest education is university degree
0.121 (0.327)
0.149 (0.356)
-0.028 (0.025)
0.103 (0.305)
0.144 (0.352)
-0.041 (0.029)
Mother’s highest education is university degree
0.125 (0.332)
0.144 (0.351)
-0.019 (0.025)
0.166 (0.373)
0.114 (0.317)
0.052* (0.027)
Number of observations 192 1,438 152 905 Sampling weights used; Standard deviations/errors are in brackets *** statistically significant at 1% ** statistically significant at 5% * statistically significant at 10%
23
Table 3: Effect of the L’Aquila earthquake on academic performance: DiD estimates Dependent variable On-time graduation Drop-out Constant -0.002
(0.368) 0.050
(0.176) Earthquake (cohort of students enrolled at the time of the L’Aquila earthquake, i.e. 2007 cohort)
0.083*** (0.014)
-0.005 (0.008)
Aquila (Enrolled at the University of L’Aquila)
0.040* (0.021)
0.012 (0.014)
Earthquake*Aquila -0.066** (0.032)
0.028 (0.018)
male -0.029** (0.012)
0.005 (0.007)
Age (omitted is 23 years or more) -21 years or less -0.033
(0.032) -0.050** (0.021)
-22 years -0.050*** (0.019)
-0.023* (0.013)
Upper secondary school final mark (omitted is between 60 and 69)
-between 70 and 79 0.001 (0.017)
-0.025** (0.010)
-between 80 and 89 0.027 (0.018)
-0.046*** (0.011)
-between 90 and 100 0.082*** (0.018)
-0.037*** (0.011)
Lower secondary school final mark (omitted is pass)
-good 0.005 (0.020)
-0.011 (0.013)
-very good 0.018 (0.021)
-0.021 (0.013)
-excellent 0.054** (0.023)
-0.017 (0.014)
Academic upper secondary school (liceo)
-0.011 (0.013)
-0.019** (0.008)
Private upper secondary school 0.025 (0.035)
0.035* (0.020)
Father’s highest education is university degree
-0.001 (0.017)
0.011 (0.011)
Mother’s highest education is university degree
-0.008 (0.018)
-0.014 (0.011)
Unemployment rate -0.001 (0.001)
0.001 (0.001)
Dummies for region of residence Yes Yes R squared 0.074 0.045 Number of observations 2,570 2,041 Sampling weights used; Standard errors are in brackets *** statistically significant at 1% ** statistically significant at 5% * statistically significant at 10%
24
Table 4: DiD estimates on on-time graduation- Checks Dependent variable Parallel trend
assumption Placebo test
Constant -0.061 (0.724)
-0.007 (0.371)
2004 student cohort 0.061*** (0.015)
0.014 (0.014)
Aquila (Enrolled at the University of L’Aquila)
0.036 (0.028)
0.020 (0.022)
2004 student cohort*Aquila -0.041 (0.037)
-0.008 (0.034)
Controls Yes Yes R squared 0.067 0.061 Number of observations 1,630 2,570 Sampling weights used; Standard errors are in brackets *** statistically significant at 1% ** statistically significant at 5% * statistically significant at 10%
25
Table 5: Effect of the L’Aquila earthquake on academic performance: DiD estimates by gender On-time graduation Drop-out Dependent variable Women Men Women Men Constant -0.021
(0.388) 1.042*** (0.396)
0.022 (0.148)
0.100 (0.262)
Earthquake (cohort of students enrolled at the time of the L’Aquila earthquake, i.e. 2007 cohort)
0.108*** (0.019)
0.043** (0.021)
-0.003 (0.008)
-0.011 (0.016)
Aquila (Enrolled at the University of L’Aquila)
0.068** (0.028)
-0.006 (0.032)
0.007 (0.014)
0.025 (0.029)
Earthquake*Aquila -0.119** (0.048)
-0.007 (0.047)
0.022 (0.020)
0.039 (0.036)
Controls Yes Yes Yes Yes R squared 0.082 0.114 0.041 0.081 Number of observations 1,647 923 1,307 734 Sampling weights used; Standard errors are in brackets *** statistically significant at 1% ** statistically significant at 5% * statistically significant at 10%
